Latest / Daniel and Kelly’s Extraordinary Universe / Why do scientists do simulations?

Transcript

Why do scientists do simulations? · Daniel and Kelly’s Extraordinary Universe

• 0:08
  Speaker 1: Hey, am I speaking to the real Joorge today?
• 0:11
  Speaker 2: Who else could it be?
• 0:12
  Speaker 1: I don't know. It could be the simulated Jorge or
• 0:14
  Speaker 1: an AI generated horheage GPT.
• 0:18
  Speaker 2: Yes, chat Jorge. Would it make a difference.
• 0:22
  Speaker 1: That sounds like something the simulated Jorge would say?
• 0:24
  Speaker 2: Mmm. I kind of wish I had simulated Horges. Then
• 0:27
  Speaker 2: I might avoid a lot of mistakes I make. Or
• 0:30
  Speaker 2: they could do all the work while I sleep in.
• 0:32
  Speaker 1: Why do you think simulated Hojoges are less likely to
• 0:34
  Speaker 1: make mistakes?
• 0:35
  Speaker 2: No? I mean they would do the mistakes, and then
• 0:36
  Speaker 2: I would learn from them. That's the idea.
• 0:39
  Speaker 3: Right.
• 0:39
  Speaker 1: That does sound useful, But I think you have to
• 0:41
  Speaker 1: be careful about the sim Jorges organizing and rising up
• 0:44
  Speaker 1: against you.
• 0:44
  Speaker 3: Oh do you think.
• 0:45
  Speaker 2: They would form their own union or a revolt?
• 0:48
  Speaker 1: Do you mean, yeah, either mutiny or fair wages. Either one.
• 0:52
  Speaker 2: Well, I could just pay them in simulated money.
• 0:53
  Speaker 1: I guess as long as they can use that to
• 0:55
  Speaker 1: feed their simulated children, I bet they'd be happy. Oh.
• 0:57
  Speaker 2: No, I definitely provide simulated benefits too. The whole simulated
• 1:01
  Speaker 2: family gets a bonus.
• 1:03
  Speaker 1: You're not a good employer, but you can simulate one.
• 1:20
  Speaker 2: Hi, I'm Jory mccartoonists and the author of Oliver's Great
• 1:23
  Speaker 2: Big Universe. Hi.
• 1:24
  Speaker 3: I'm Daniel.
• 1:25
  Speaker 1: I'm a particle physicist and a professor at UC Irvine,
• 1:28
  Speaker 1: and I am constantly simulating crazy conditions.
• 1:32
  Speaker 2: You mean in your life or in your work.
• 1:35
  Speaker 1: Well, work is a big part of my life. But yeah,
• 1:37
  Speaker 1: my job involves simulating collisions at very high energies all
• 1:41
  Speaker 1: the time.
• 1:42
  Speaker 2: And you also actually do them, right, You actually collide
• 1:45
  Speaker 2: things at the Large Hadron Collider.
• 1:46
  Speaker 1: That's right. We both collide particles together in real life,
• 1:50
  Speaker 1: and we simulate what would happen if we collided particles
• 1:53
  Speaker 1: under various different potential laws of the universe to see
• 1:56
  Speaker 1: what might happen. Will the Earth get gobbled up with
• 2:00
  Speaker 1: to create a black hole that destroys the Earth or not?
• 2:02
  Speaker 1: Let's find out.
• 2:03
  Speaker 2: Does that mean your whole career is a simulation or
• 2:07
  Speaker 2: your whole life?
• 2:08
  Speaker 1: You know, our computers are not fast enough to keep
• 2:10
  Speaker 1: up with reality. So while we generate lots and lots
• 2:13
  Speaker 1: of simulated collisions, the real collider has generated more collisions
• 2:17
  Speaker 1: than we could ever simulate.
• 2:19
  Speaker 2: But how do you know, Daniel, that we're not in
• 2:21
  Speaker 2: a simulation right now, Like you might think you're doing experiments,
• 2:24
  Speaker 2: but really you're just inside of a video game somewhere.
• 2:26
  Speaker 1: Well, I want to find the cheap goods, then, well I.
• 2:28
  Speaker 2: Think if you had found them by now, you probably
• 2:30
  Speaker 2: would have a noble price.
• 2:31
  Speaker 1: Right, I'm hoping smashing particles together gives me the cheek goods.
• 2:34
  Speaker 2: And then that opens up the real boss level.
• 2:38
  Speaker 1: Where I fight the simulated army of Jorges.
• 2:40
  Speaker 2: No, you fight the real Jegeo. That's the real boss.
• 2:45
  Speaker 2: But anyways, welcome to our podcast Daniel and Jorge Explain
• 2:47
  Speaker 2: the Universe, a production of iHeartRadio.
• 2:50
  Speaker 1: In which we use our tiny little minds to try
• 2:52
  Speaker 1: to understand the vast universe. We hope to build in
• 2:55
  Speaker 1: your head a simulation of sorts, one that describes the
• 2:59
  Speaker 1: way the real universe works out there. We hope to
• 3:03
  Speaker 1: encode into your brain some laws of physics that will
• 3:06
  Speaker 1: help you understand how the real universe out there is
• 3:09
  Speaker 1: smashing and bashing to create our Bonker's reality.
• 3:12
  Speaker 2: That's right, the universe are doing all kinds of amazing
• 3:15
  Speaker 2: and awe sometimes and saying things out there in reality.
• 3:18
  Speaker 2: And so it's our job as humans and as scientists
• 3:22
  Speaker 2: to understand what's going on and to ask questions, to
• 3:25
  Speaker 2: probe into the true answers to why things are the
• 3:27
  Speaker 2: way they are.
• 3:28
  Speaker 1: And the classical way that science does this is with
• 3:30
  Speaker 1: theories and experiments, hypotheses and tests. You have an idea,
• 3:35
  Speaker 1: you go and see out there in the universe. If
• 3:37
  Speaker 1: it works, you predict something happens, and you go and
• 3:40
  Speaker 1: check to see if it does. But the modern scientific
• 3:43
  Speaker 1: method has a third way, which lives sort of uncomfortably
• 3:45
  Speaker 1: between theory and experiments.
• 3:48
  Speaker 2: Oh, why is it uncomfortable? Is it like uncomfortable, awkward
• 3:51
  Speaker 2: or uncomfortable, like physically uncomfortable.
• 3:53
  Speaker 1: It's a little bit uncomfortable for those of us who
• 3:55
  Speaker 1: specialize in not to know where we fit into the
• 3:57
  Speaker 1: picture of science. Some people consider me experimental physicists. Some
• 4:01
  Speaker 1: people are like, nah, he mostly runs simulation, so he's
• 4:03
  Speaker 1: really a theorist. So you can be sort of like
• 4:06
  Speaker 1: uncomfortably between two different communities. If you do a lot
• 4:09
  Speaker 1: of simulations.
• 4:10
  Speaker 2: You start your own simulated community. Can you be like
• 4:13
  Speaker 2: a simulating physicist.
• 4:15
  Speaker 1: A simulator a simulate trist. That sounds not safe for work.
• 4:20
  Speaker 2: Actually, I don't know what you mean, but I'll take
• 4:24
  Speaker 2: your word for it.
• 4:25
  Speaker 1: You know, academic communities change pretty slowly, and for example,
• 4:28
  Speaker 1: in departments of physics, people tend to hire people that
• 4:32
  Speaker 1: are like them. The experimentalists get to hire somebody. They
• 4:35
  Speaker 1: want to hire somebody who's a blue blooded experimentalist down
• 4:38
  Speaker 1: to the core. So if you work at the intersection
• 4:41
  Speaker 1: of fields, you do some experiments, you do some theory,
• 4:44
  Speaker 1: maybe even do some computer science and machine learning, then
• 4:47
  Speaker 1: you don't necessarily have a home, you don't have a
• 4:48
  Speaker 1: tribe that's going to go to bat for you to
• 4:50
  Speaker 1: get hired. So it's sort of about the sociology of
• 4:52
  Speaker 1: science as a real practice.
• 4:54
  Speaker 2: Well, I think that kind of makes sense, right, Like
• 4:56
  Speaker 2: why hire a simulating physicist when you can just simulate one?
• 5:00
  Speaker 2: Why go to all the trouble?
• 5:02
  Speaker 1: You know, I think that's true. If we could simulate physicists,
• 5:04
  Speaker 1: we could get a lot more done. But we're not
• 5:06
  Speaker 1: quite there yet. Human physicists still have a little bit
• 5:09
  Speaker 1: of an edge.
• 5:10
  Speaker 2: Yeah, maybe till next week when chat GPT catches up
• 5:14
  Speaker 2: and starts doing physics.
• 5:16
  Speaker 1: I mean, have you ever asked chat GPT a physics question.
• 5:18
  Speaker 1: You don't get physics out, that's for sure.
• 5:21
  Speaker 2: But anyways, it is an interesting universe because I guess
• 5:23
  Speaker 2: sometimes there are questions. You can't just go out there
• 5:26
  Speaker 2: and try for yourself in the universe, right, that's right.
• 5:28
  Speaker 1: Simulation has emerged in the past fifty years as an
• 5:31
  Speaker 1: extraordinarily powerful tool as a little bit of experiment and
• 5:34
  Speaker 1: a little bit of theory, and it lets us answer
• 5:37
  Speaker 1: questions that we otherwise could not answer. It really is
• 5:40
  Speaker 1: a completely new tool in the science tool.
• 5:43
  Speaker 2: Belt, although I would argue it's maybe one of the
• 5:45
  Speaker 2: oldest tools in science. And so today on the podcast,
• 5:48
  Speaker 2: we'll be asking the question why do scientists do simulations?
• 5:58
  Speaker 2: Why do scientists do anything.
• 6:00
  Speaker 1: Other than the obvious that simulations are so much fun.
• 6:03
  Speaker 1: You get to build your own little universe. You are
• 6:05
  Speaker 1: the creator and god of that simulated universe.
• 6:08
  Speaker 2: Oh boy, is that that the ultimate goal? There? To
• 6:12
  Speaker 2: be gods? No, you know, you don't have to get
• 6:17
  Speaker 2: a degree for that. You could just buy some legos.
• 6:20
  Speaker 1: Oh, I've been doing that since I was a little kid.
• 6:22
  Speaker 1: I just want more and more powerful simulations.
• 6:26
  Speaker 2: You want more powerful legos, smaller legos.
• 6:28
  Speaker 1: Jokes aside, There is a real sense of power when
• 6:31
  Speaker 1: you create a simulated universe, because you are deciding what
• 6:34
  Speaker 1: the laws of physics are in that universe, what particles
• 6:37
  Speaker 1: do they have, how do they interact? And then you
• 6:39
  Speaker 1: get to see how it all plays out.
• 6:42
  Speaker 2: Mmmm yeah, I sort of get that. I mean I
• 6:44
  Speaker 2: write a lot, I create characters, and I sort of
• 6:46
  Speaker 2: build my own world. What's the difference.
• 6:49
  Speaker 1: Yeah, you could think of fiction as simulated human interaction
• 6:53
  Speaker 1: and lives. Right, we're exploring what it would be like
• 6:55
  Speaker 1: to be in those situations.
• 6:58
  Speaker 2: Yeah. Wait, did you just say your work.
• 6:59
  Speaker 1: Is simulations are definitely fiction. Sometimes they align with reality,
• 7:04
  Speaker 1: and one deep question is how well they align? What
• 7:07
  Speaker 1: lessons you can learn from your simulated fiction that carry
• 7:11
  Speaker 1: over into the real world.
• 7:13
  Speaker 2: Interesting, So your research is science fiction is what you're saying.
• 7:20
  Speaker 1: You know, I've always argued that there's a strong connection
• 7:23
  Speaker 1: between science and science fiction.
• 7:25
  Speaker 3: Right.
• 7:25
  Speaker 1: One aspect of science is like, well, what are the laws?
• 7:27
  Speaker 3: Could they be?
• 7:28
  Speaker 1: This?
• 7:28
  Speaker 3: Could they be?
• 7:28
  Speaker 2: That?
• 7:29
  Speaker 1: There's an element of creativity and exploration there, absolutely so. Yeah,
• 7:33
  Speaker 1: I'm constantly creating science fiction universes and trying to see
• 7:36
  Speaker 1: if they line up with ours.
• 7:39
  Speaker 2: And then you wonder why the other physicis don't want
• 7:40
  Speaker 2: to play with you.
• 7:43
  Speaker 1: Fortunately I got tenured before I revealed all of these
• 7:46
  Speaker 1: crazy instincts. That's the game.
• 7:50
  Speaker 2: Well, this is an interesting question, and so as usual,
• 7:52
  Speaker 2: we were wondering how many people out there had thought
• 7:55
  Speaker 2: about why scientists do the things they do, and in particular,
• 7:58
  Speaker 2: why they do simulations in their work.
• 8:00
  Speaker 1: So thanks very much to everybody who answers these questions
• 8:03
  Speaker 1: for this fun segment of the podcast, one of my favorites.
• 8:06
  Speaker 1: If you like to join the team or just answer
• 8:09
  Speaker 1: one or two questions right to us to questions at
• 8:12
  Speaker 1: Danielandhorge dot com.
• 8:14
  Speaker 2: So think about it for a second. If someone asks
• 8:16
  Speaker 2: you why scientists do simulations, what would you say.
• 8:19
  Speaker 4: Well, using simulations, we can observe scenarios in our models
• 8:26
  Speaker 4: that we can't necessarily observe in real life, and see
• 8:32
  Speaker 4: what can happen in certain situations like in a black
• 8:36
  Speaker 4: hole or when galaxies collide or something like that.
• 8:38
  Speaker 5: Scientists do simulations because the universe is really old, really big,
• 8:45
  Speaker 5: sometimes really destructive. Frankly, I'm happy they do a lot
• 8:49
  Speaker 5: of that modeling and simulations and don't necessarily try to
• 8:52
  Speaker 5: create big bang conditions on a big scale, or the
• 8:57
  Speaker 5: explosion of stars or something.
• 9:00
  Speaker 2: Couple of interesting answers did anyone look at you funny
• 9:03
  Speaker 2: when you ask them the question.
• 9:05
  Speaker 1: I don't know. These were all on the internet, so
• 9:08
  Speaker 1: I couldn't capture their facial expressions. Did they send an emoji?
• 9:12
  Speaker 1: But I do sense some relief in there that, for example,
• 9:15
  Speaker 1: we are trying to simulate galaxy collisions rather than trying
• 9:18
  Speaker 1: to arrange galaxy collisions.
• 9:20
  Speaker 2: Well, if we could do that, that'd be pretty cool.
• 9:23
  Speaker 2: I mean, not for those galaxies, but just to have
• 9:25
  Speaker 2: that power.
• 9:26
  Speaker 1: Yeah, you'd have to have like sign offs from every
• 9:28
  Speaker 1: alien civilization in both galaxies before you could even begin.
• 9:32
  Speaker 2: I guess that would be the polite thing to do. Yes.
• 9:35
  Speaker 2: But anyways, as you were saying, this is a big
• 9:37
  Speaker 2: part of how science is done these days, and so Daniel,
• 9:40
  Speaker 2: I guess let's start from the basics. What is a
• 9:43
  Speaker 2: simulation in your view as a physicists.
• 9:45
  Speaker 1: So, a simulation, or more specifically, a computer simulation, is
• 9:48
  Speaker 1: a specific program that involves a scientific model. A model
• 9:53
  Speaker 1: is like our picture of how the world might work.
• 9:56
  Speaker 1: It's like a simplified version of the real universe. Let's
• 10:00
  Speaker 1: us explore a specific question, and a simulation is usually
• 10:04
  Speaker 1: a program on a computer that uses like step by
• 10:07
  Speaker 1: step methods to explore the behavior of that model.
• 10:11
  Speaker 2: And usually this model has the form of an equation, right, Like,
• 10:15
  Speaker 2: for example, F equalsma is a model of the world, right,
• 10:18
  Speaker 2: and how things move in the world.
• 10:20
  Speaker 1: Yeah, the science we do is mathematical, and the way
• 10:22
  Speaker 1: we describe things is mathematical, and so usually that involves equations,
• 10:26
  Speaker 1: equations that represent constraints on the model, like the way
• 10:29
  Speaker 1: things have to happen. And as you say, F equals
• 10:32
  Speaker 1: ma is a model. If I want to toss a
• 10:34
  Speaker 1: baseball across my backyard, I want to answer the question
• 10:37
  Speaker 1: where is it going to land? Then I have lots
• 10:39
  Speaker 1: of possible ways to answer that question, but the most
• 10:42
  Speaker 1: appropriate ways to make the simplest model possible that still
• 10:46
  Speaker 1: captures everything that I'm interested in, and so often in
• 10:50
  Speaker 1: our world, like when we're tossing baseballs, we can do
• 10:53
  Speaker 1: something pretty simple just F equals ma, which ignores all
• 10:56
  Speaker 1: sorts of swarming quantum details about what's happening inside at
• 11:00
  Speaker 1: the baseball and just describes simple motion of a parabole.
• 11:04
  Speaker 2: Yeah, it's almost like you. I mean, as a scientist,
• 11:07
  Speaker 2: you're trying to come up with the rules of the universe, right,
• 11:09
  Speaker 2: that's sort of the goal of science, right, And what
• 11:11
  Speaker 2: sometimes that rule looks like is in a question that says,
• 11:15
  Speaker 2: you know, if you have a mass and you apply
• 11:17
  Speaker 2: a force to it, then it's going to start moving
• 11:19
  Speaker 2: with a certain acceleration.
• 11:21
  Speaker 1: Exactly in your words, these are all science fictions. We're
• 11:23
  Speaker 1: living in this world and we're wondering what are the rules,
• 11:26
  Speaker 1: and so we're trying a bunch of different rules, saying
• 11:28
  Speaker 1: does this rule describe our universe? Does that rule describe
• 11:30
  Speaker 1: our universe? So every sort of theoretical exploration of the
• 11:34
  Speaker 1: universe involves building a model and then asking the question
• 11:37
  Speaker 1: does that model align with the reality that we see.
• 11:41
  Speaker 1: Computer simulations are a special kind of model or a
• 11:43
  Speaker 1: special what a test? Really complicated models that we can't
• 11:47
  Speaker 1: otherwise test, Like the model F equals M a pretty simple.
• 11:50
  Speaker 1: I can use pencil and paper to make predictions, and
• 11:53
  Speaker 1: then I can throw a ball in my backyard to
• 11:55
  Speaker 1: confirm those predictions.
• 11:56
  Speaker 2: Right, because I guess F equals A has like a
• 11:58
  Speaker 2: mathematical But I think the idea is that you take
• 12:01
  Speaker 2: in a question like F equals in May and you
• 12:03
  Speaker 2: basically program that into a computer and say, you know,
• 12:06
  Speaker 2: any masses in this program, they have to move according
• 12:09
  Speaker 2: to this law.
• 12:10
  Speaker 1: That's right. If, for example, I don't just want to
• 12:13
  Speaker 1: describe one ball, but I want to describe like ten
• 12:15
  Speaker 1: to the twenty five balls, right, some huge number of balls.
• 12:19
  Speaker 1: Maybe I'm modeling an ideal gas, or like a swimming
• 12:22
  Speaker 1: pool full of ping pong balls or something, and I
• 12:25
  Speaker 1: want to describe that. Then I can no longer use
• 12:27
  Speaker 1: pencil and paper. But you're in. I can take those
• 12:29
  Speaker 1: equations and put them into a computer and ask the
• 12:31
  Speaker 1: computer to force those balls to follow that equation, and
• 12:34
  Speaker 1: then I could see what happens. It's sort of like
• 12:36
  Speaker 1: a virtual experiment.
• 12:38
  Speaker 2: Yeah, it's like you're creating your own little universe.
• 12:41
  Speaker 1: Right exactly. And this becomes super essential when we don't
• 12:44
  Speaker 1: have like a single equation that describes everything, Like we
• 12:48
  Speaker 1: don't have a solution to what happens when you put
• 12:51
  Speaker 1: ten to the twenty five ping pong balls into a
• 12:53
  Speaker 1: swimming pool. We just don't know how to do that
• 12:56
  Speaker 1: calculation to come up with some nice summary of the results.
• 13:00
  Speaker 1: What we can do is put it into a computer
• 13:01
  Speaker 1: and have the computer step it forward in time very carefully,
• 13:05
  Speaker 1: and we can see what happens without ever actually having
• 13:07
  Speaker 1: to buy that number of ping bomb balls.
• 13:09
  Speaker 2: Right. That's sort of the power of the computer, right,
• 13:12
  Speaker 2: Like you can simulate one ball, which is a calculator, right,
• 13:16
  Speaker 2: Like you can say after one second, it's going to
• 13:18
  Speaker 2: be here, after two seconds, it is going to be
• 13:19
  Speaker 2: here by following these rules. But if you have, like
• 13:22
  Speaker 2: you said, a whole bunch of balls, or a more
• 13:26
  Speaker 2: complicated system, then a computer can sort of do all
• 13:29
  Speaker 2: those calculations for you faster, exactly.
• 13:31
  Speaker 1: One huge advantage is tackling a very large number of objects,
• 13:35
  Speaker 1: and the other is when we don't have the equations,
• 13:38
  Speaker 1: we don't know how to solve them, Like for f equals,
• 13:40
  Speaker 1: I may we know how to solve that. Technically, that
• 13:42
  Speaker 1: is a differential equation because A is a second derivative
• 13:46
  Speaker 1: of position, right, there's derivatives on both sides. And in
• 13:50
  Speaker 1: general and mathematics, differential equations are very very hard to solve.
• 13:53
  Speaker 1: This a small number that we actually know how to solve.
• 13:56
  Speaker 1: So sometimes you have a system that's described by a
• 13:58
  Speaker 1: differential equation you don't know how to solve, Like fluid flow,
• 14:02
  Speaker 1: for example, described by the Navier Stokes equation. We don't
• 14:05
  Speaker 1: know how to solve that. In general. But what we
• 14:07
  Speaker 1: can do on a computer is approximated. You can say,
• 14:10
  Speaker 1: you know, let's just move it forward in time, not
• 14:12
  Speaker 1: a year or a minute or some long period of time,
• 14:15
  Speaker 1: but just like a microsecond, and across a microsecond, we
• 14:18
  Speaker 1: can make some approximations. We can say, let's not use
• 14:21
  Speaker 1: the full equation, let's simplify it and take some like
• 14:23
  Speaker 1: linear approximation of it, and then if we take a
• 14:25
  Speaker 1: lot of tiny little steps, we hope that we roughly
• 14:28
  Speaker 1: get the right answer.
• 14:29
  Speaker 2: Right, because I think, as you were saying, like something
• 14:31
  Speaker 2: like f EQUOSM has the solution, meaning that you can
• 14:34
  Speaker 2: derive a formula for like the precision of your ball
• 14:37
  Speaker 2: at all times, where you can just like after three
• 14:40
  Speaker 2: seconds you just put the time in and it gives
• 14:42
  Speaker 2: you the position of the ball, right, because you can
• 14:44
  Speaker 2: integrate that equation and find the solution. But some equations
• 14:46
  Speaker 2: you can, like they're so complex you can get a
• 14:49
  Speaker 2: formally that will tell you what's going to happen ten
• 14:51
  Speaker 2: years from now or twenty years from now. Right, those
• 14:53
  Speaker 2: you need to do little steps by little steps exactly.
• 14:56
  Speaker 1: And the crucial idea there is that you're making a
• 14:58
  Speaker 1: linear approximation you're taking the full equation which you don't
• 15:00
  Speaker 1: know how to solve, and you're saying, well, let's replace
• 15:02
  Speaker 1: it with an approximate version of it, which is not
• 15:05
  Speaker 1: going to be correct, but it might be correct for
• 15:07
  Speaker 1: like a microsecond. And so we'll use the approximate linear
• 15:11
  Speaker 1: version of that that we do know how to solve
• 15:12
  Speaker 1: from a tiny little step, and then we'll start again,
• 15:15
  Speaker 1: and we'll make another tiny little step, and we hope
• 15:17
  Speaker 1: them little mistakes cancel out and don't build up into
• 15:19
  Speaker 1: some big overall mistake.
• 15:21
  Speaker 2: Right Yeah. It's almost like if you take small enough steps,
• 15:24
  Speaker 2: then you're less likely to deviate from the reality of.
• 15:27
  Speaker 1: It, right, Yeah, exactly. And people who do approximations know
• 15:30
  Speaker 1: that there's lots of times this is useful. Like you
• 15:32
  Speaker 1: want to calculate Trigg function like sign sign is really
• 15:35
  Speaker 1: hard to calculate like sort of from scratch, but for
• 15:38
  Speaker 1: very small values of the angle, sign of x is
• 15:41
  Speaker 1: just equal to x, you can like approximate this complicated
• 15:44
  Speaker 1: function with a simple one. It mostly gets the right answer.
• 15:47
  Speaker 1: That's just one example.
• 15:48
  Speaker 2: I'm not sure you're going to Trigonometry usually makes things
• 15:51
  Speaker 2: you understand, but I think I think we get the idea,
• 15:55
  Speaker 2: which is that you know, if you take small enough
• 15:57
  Speaker 2: steps and you you sort of a simplified version of
• 16:00
  Speaker 2: your model, then you're less likely to make mistakes.
• 16:04
  Speaker 1: Exactly. You can't trust those approximations forward a second or
• 16:07
  Speaker 1: a minute or a year, but you could trust them
• 16:09
  Speaker 1: like a microsecond. And so you have the simulated universe
• 16:13
  Speaker 1: in your computer. You feed in the initial conditions, and
• 16:16
  Speaker 1: then you ask it to take a step forward in time,
• 16:18
  Speaker 1: and you ask it to take another step forward, and
• 16:20
  Speaker 1: if you have enough computing power, you can run it
• 16:22
  Speaker 1: for a while and you can see what happens to
• 16:25
  Speaker 1: all my ping pong balls in my simulated swimming pool,
• 16:28
  Speaker 1: or what happens to my galaxy as the stars all
• 16:30
  Speaker 1: swirl around each other.
• 16:32
  Speaker 2: Right, Like you were saying, like fluids are notoriously really
• 16:35
  Speaker 2: hard to solve as an equation, right, These are really
• 16:38
  Speaker 2: complex equations that govern what's going on because they sort
• 16:41
  Speaker 2: of like depend on a lot of things. Like there's
• 16:43
  Speaker 2: a lot going on, right, there's time, and then there's
• 16:46
  Speaker 2: distance and the velocity of things and all those factor
• 16:49
  Speaker 2: in that's hard or impossible to like predict exactly what's
• 16:52
  Speaker 2: going to happen in the future exactly.
• 16:54
  Speaker 1: And the big complication there is the interactions. Like back
• 16:57
  Speaker 1: to the ping pong balls. If you just had a
• 16:58
  Speaker 1: lot of ping pong balls and they're all flying around
• 17:01
  Speaker 1: but not touching each other, it wouldn't be that hard
• 17:03
  Speaker 1: to calculate what's going to happen to each one, But
• 17:06
  Speaker 1: as soon as they start banging against each other, becomes
• 17:08
  Speaker 1: much much more complicated because the solution of ping pong
• 17:11
  Speaker 1: ball number six hundred and forty two now depends on
• 17:13
  Speaker 1: ping pong ball number one, one hundred and eleven and
• 17:17
  Speaker 1: every other ping pong ball, so becomes much more complicated.
• 17:20
  Speaker 1: And that's why fluids are so complicated, because every sheet
• 17:23
  Speaker 1: of the fluid depends on the friction with the other
• 17:25
  Speaker 1: sheet of the fluid. And that's what makes the Navier
• 17:28
  Speaker 1: Stokes equation, for example, so intractable.
• 17:30
  Speaker 2: Right, And so you take it little by little, and
• 17:32
  Speaker 2: so you say, okay, this time, I'm going to ignore
• 17:34
  Speaker 2: some of these effects and just take one small step
• 17:37
  Speaker 2: to see where all those little molecules go. And then
• 17:40
  Speaker 2: you keep repeating that and hopefully it sort of looks
• 17:44
  Speaker 2: like the real thing.
• 17:45
  Speaker 1: Exactly, and it gives you this incredible power that you
• 17:48
  Speaker 1: can hopefully identify emergent behavior. The way we do science
• 17:52
  Speaker 1: in our universe is that we like focus on one
• 17:54
  Speaker 1: level where we understand things we can describe, like the
• 17:57
  Speaker 1: microphysics of how particles being against each other. But sometimes
• 18:00
  Speaker 1: we're interested in things at another level, Like you understand
• 18:03
  Speaker 1: how rain drops move through the wind, but your real
• 18:05
  Speaker 1: question is like is this hurricane going to hit Florida
• 18:08
  Speaker 1: or Alabama? And so even if you don't have like
• 18:10
  Speaker 1: an equation that describes hurricanes, if you have an equation
• 18:13
  Speaker 1: that describes the rain drops, you can feed that all
• 18:16
  Speaker 1: into your computer, run simulations, and then get answers to
• 18:19
  Speaker 1: your higher level question. You can see like the emergent
• 18:22
  Speaker 1: phenomena of the hurricane in simulation.
• 18:25
  Speaker 2: Well, it's dig a little bit deeper into how simulation
• 18:28
  Speaker 2: works and the things like weather and why scientists use
• 18:32
  Speaker 2: simulations to try to learn things about the real universe.
• 18:36
  Speaker 2: But first let's take a quick break. All right, we're
• 18:51
  Speaker 2: having a simulation of a podcast here, right, We're not
• 18:53
  Speaker 2: really having a podcast, right, We're just pretending to have
• 18:56
  Speaker 2: a podcast.
• 18:56
  Speaker 1: We're simulating the process of injecting ideas into listen in our.
• 19:00
  Speaker 2: Minds, and so we're talking about why scientists use simulations,
• 19:03
  Speaker 2: and it's kind of, I guess a philosophical question, perhaps
• 19:07
  Speaker 2: because doing a simulation of reality is not really reality, right,
• 19:11
  Speaker 2: and so, and you're not really experimenting on reality. So
• 19:14
  Speaker 2: it's kind of a I guess, a funny thing for
• 19:16
  Speaker 2: scientists to do it because you're not really doing experiments
• 19:19
  Speaker 2: in the real world. But it's at the same time
• 19:21
  Speaker 2: really helpful. Right.
• 19:22
  Speaker 1: That's true, But that same criticism could be applied to
• 19:24
  Speaker 1: basically everything in science. When we do science, we never
• 19:28
  Speaker 1: use all of the full gory details of the universe
• 19:32
  Speaker 1: to answer a question. We're always using some stripped down
• 19:35
  Speaker 1: version because otherwise it's totally intractable. Like when we do
• 19:39
  Speaker 1: F equals M, even for a single ball flying through
• 19:42
  Speaker 1: the air, we're ignoring lots of stuff. We're ignoring air resistance,
• 19:45
  Speaker 1: we're ignoring quantum effects of the particles inside of it.
• 19:48
  Speaker 1: We're ignoring all sorts of things because we don't think
• 19:50
  Speaker 1: that they are important. And so every time you build
• 19:53
  Speaker 1: a model of the universe, theoretical or simulation, you're always
• 19:56
  Speaker 1: making a choice about what to ignore and what to include.
• 19:59
  Speaker 2: Well, we talked a lot about what a simulation is right.
• 20:02
  Speaker 2: It's a computer program where you program in the rules
• 20:04
  Speaker 2: that you think that the world follows the rules of
• 20:07
  Speaker 2: the universe, at least in your simulated universe, and then
• 20:10
  Speaker 2: you sort of let the computer kind of run this world,
• 20:13
  Speaker 2: and then it sort of tells you what may or
• 20:15
  Speaker 2: may will sort of happen.
• 20:17
  Speaker 1: Exactly, and it lets you examine all sorts of universes
• 20:20
  Speaker 1: you don't otherwise have access to, like in my work,
• 20:23
  Speaker 1: and let's me answer questions like what would I see
• 20:25
  Speaker 1: in our particle detectors if the Higgs boson was this
• 20:29
  Speaker 1: kind of particle, or what if there was no Higgs boson,
• 20:32
  Speaker 1: or what if it had twice the mass that it had?
• 20:34
  Speaker 1: What would we see in our detectors? What would that
• 20:36
  Speaker 1: universe be?
• 20:37
  Speaker 2: Like?
• 20:37
  Speaker 1: What would those experiments result in?
• 20:39
  Speaker 2: So it gives you ideas for experiments, or it's a
• 20:42
  Speaker 2: sort it can guide your real experiments. Right, that's part
• 20:45
  Speaker 2: of the idea.
• 20:46
  Speaker 1: Right, It's actually crucial for interpreting our experiments. When we
• 20:49
  Speaker 1: look at data from the actual collider and we see
• 20:52
  Speaker 1: these splashes of energy here and splashes of energy there,
• 20:54
  Speaker 1: and we look at the patterns the correlations. The way
• 20:57
  Speaker 1: we interpret those is by comparing them to simulations, we say,
• 21:01
  Speaker 1: is this consistent with the higgs boson with these properties
• 21:04
  Speaker 1: or is it more consistent with the higgs boson with
• 21:06
  Speaker 1: some other properties. So the simulation in some sense defines
• 21:09
  Speaker 1: the ideas that we're considering, the various hypotheses that we're
• 21:12
  Speaker 1: trying to distinguish between.
• 21:14
  Speaker 2: Right, It lets you explore the possibilities. That's the idea
• 21:17
  Speaker 2: of a simulation. Right, Let's you maybe make mistakes.
• 21:19
  Speaker 1: Absolutely, and before we build a detector, we simulated to
• 21:22
  Speaker 1: see like is this going to work or how well
• 21:24
  Speaker 1: is it going to perform? Or oops, turns out we
• 21:27
  Speaker 1: need to swap the order these two things or nothing's
• 21:29
  Speaker 1: going to work. So yeah, making mistakes and simulation is
• 21:31
  Speaker 1: much cheaper than making them in reality.
• 21:33
  Speaker 2: Yeah, And as you were saying, simulations play a big
• 21:36
  Speaker 2: part in weather prediction, right, I mean that's how weather
• 21:38
  Speaker 2: predictions work. Like when you look at the weather forecast
• 21:41
  Speaker 2: and says it's going to rain tomorrow, it's because some
• 21:43
  Speaker 2: big computer out there has basically taken the data from
• 21:46
  Speaker 2: today and simulated what's going to happen tomorrow exactly.
• 21:49
  Speaker 1: And that's really the origin of computer simulations. People wanted
• 21:54
  Speaker 1: to predict the weather, to understand what's going to happen
• 21:57
  Speaker 1: to these cloud patterns. But nobody could really do it
• 21:59
  Speaker 1: with pen and papers. Too complicated, too many pieces of information,
• 22:02
  Speaker 1: and the equations are really just a mess. So meteorology
• 22:06
  Speaker 1: is one of the first places where people decided, let's
• 22:08
  Speaker 1: code this up on the computer to try to grapple
• 22:10
  Speaker 1: with this complexity and see if we can get anything
• 22:13
  Speaker 1: right now, just after World War Two when computers were
• 22:15
  Speaker 1: first displaying like computational power, and it was the weather
• 22:18
  Speaker 1: forecasters and the nuclear physicists that first really jumped on
• 22:22
  Speaker 1: this train.
• 22:23
  Speaker 2: Yeah, because I think the way the weather works is
• 22:25
  Speaker 2: that you have all this data mateiological data, weather data
• 22:30
  Speaker 2: across let's say the United States, that tells you the
• 22:33
  Speaker 2: wind speeds and the clouds and the pressures and all that,
• 22:36
  Speaker 2: and then you can use it and put it into
• 22:38
  Speaker 2: basically your computer, which has a model of what should
• 22:41
  Speaker 2: happen next if that's you have all these pressures and
• 22:45
  Speaker 2: wind patterns exactly.
• 22:47
  Speaker 1: And those models are not perfect, they don't describe everything,
• 22:50
  Speaker 1: and so they're most reliable over short times because that's
• 22:53
  Speaker 1: when the errors are not going to compound as much,
• 22:55
  Speaker 1: which is why I like the prediction for how hot
• 22:57
  Speaker 1: it's going to be tomorrow is much more reliable than
• 22:59
  Speaker 1: the prediction for how hot it's going to be in
• 23:01
  Speaker 1: ten years, which you basically have no information about. Or
• 23:05
  Speaker 1: if you look at those projections for like where's the
• 23:07
  Speaker 1: hurricane going to be, the potential path of the hurricane
• 23:10
  Speaker 1: gets wider as the prediction gets further out because there's
• 23:13
  Speaker 1: more uncertainty, like is it going to hit Alabama? We
• 23:16
  Speaker 1: don't know.
• 23:16
  Speaker 2: Yeah, it's pretty cool and actually an interesting fact. I
• 23:18
  Speaker 2: would just talk to a hurricane scientist a couple of
• 23:22
  Speaker 2: months ago, and he was saying that we're still at
• 23:25
  Speaker 2: the point apparently where humans outperform simulations, even supercomputer simulations.
• 23:30
  Speaker 1: Humans using pencil and paper, or just humans like with
• 23:33
  Speaker 1: their intuition.
• 23:34
  Speaker 2: Humans with their intuition. So like, apparently we're still at
• 23:36
  Speaker 2: the point where if if you're seeing a hurricane move,
• 23:39
  Speaker 2: you run a computer simulation about where it's going to
• 23:41
  Speaker 2: go next. A human or like a season experienced hurricane
• 23:45
  Speaker 2: watcher will still today better at predicting what the hurricane
• 23:48
  Speaker 2: is going to do, just from like what's going on
• 23:50
  Speaker 2: inside their brain and the history of what they've seen before.
• 23:53
  Speaker 2: But it's getting apparently closer and closer, so maybe in
• 23:56
  Speaker 2: the near future computers will make those hurricane watchers. Is
• 24:00
  Speaker 2: totally obsleete.
• 24:01
  Speaker 1: That's super fascinating and it's fun to think about what's
• 24:03
  Speaker 1: going on inside that person's brain. They have built in
• 24:07
  Speaker 1: their head some neural network with literal biological neurons, right
• 24:11
  Speaker 1: and not your typical artificial neural network that models hurricanes,
• 24:14
  Speaker 1: and they've trained it on a bunch of real hurricanes.
• 24:16
  Speaker 1: So that's pretty cool.
• 24:18
  Speaker 2: Yeah, it makes you wonder if maybe, like in the future,
• 24:20
  Speaker 2: they're going to use AIS to predict the weather, maybe
• 24:23
  Speaker 2: you don't need a scientific model of what's going on.
• 24:26
  Speaker 1: That's a really fascinating question because AIS are already being
• 24:29
  Speaker 1: used to help boost simulations. One problem with simulations is
• 24:33
  Speaker 1: that they can be very expensive computationally. You have lots
• 24:37
  Speaker 1: and lots of rain jops and you want to model
• 24:39
  Speaker 1: it very, very accurately. It takes a computer a long
• 24:41
  Speaker 1: time to calculate every rain job and move it forward
• 24:44
  Speaker 1: in time. You want to predict something a few days out,
• 24:47
  Speaker 1: it can be very expensive computationally. We run into this
• 24:49
  Speaker 1: problem in particle physics all the time because we want
• 24:51
  Speaker 1: to simulate billions and billions of potential collisions, and the
• 24:54
  Speaker 1: interactions with the detector are very complicated. So to generate
• 24:58
  Speaker 1: one simulated collision, for example, like thirty minutes, even on
• 25:01
  Speaker 1: a modern computer, we use AI to boost those to
• 25:04
  Speaker 1: make them faster. Essentially, we train machine learning algorithms to
• 25:08
  Speaker 1: reproduce what the careful calculations have done. They don't understand it,
• 25:13
  Speaker 1: they don't like have the same equations built in. It's
• 25:15
  Speaker 1: just sort of like those people watching the examples and
• 25:18
  Speaker 1: getting an intuition. This is like a machine learning intuition.
• 25:22
  Speaker 2: So now you're not just similar working in a Meida world.
• 25:26
  Speaker 2: Now you're in twitting your way through a Meida world.
• 25:29
  Speaker 1: Yeah. And one problem is that we don't always know
• 25:31
  Speaker 1: if their predictions are accurate or why they make them.
• 25:34
  Speaker 1: You can't ask them like why did this go left
• 25:36
  Speaker 1: instead of right? They just have an internal model, the
• 25:39
  Speaker 1: same way your hurricane watchers probably can't answer detailed questions
• 25:42
  Speaker 1: about why they feel it's going this way. They just
• 25:45
  Speaker 1: feel it.
• 25:46
  Speaker 2: Yeah. And so it also raises these interesting philosophical questions
• 25:50
  Speaker 2: about what science is right, Like is it still science
• 25:52
  Speaker 2: if you get an AI to predict what's going on.
• 25:54
  Speaker 2: Even if you don't understand what the AI did.
• 25:56
  Speaker 1: It's a deep question that we're struggling with all the time.
• 25:59
  Speaker 1: But with no controversial is that it gives us extraordinary
• 26:02
  Speaker 1: power to do things we just couldn't do otherwise. We
• 26:05
  Speaker 1: can now run our simulations for much much longer and
• 26:07
  Speaker 1: in much more depth. You want to know what's going
• 26:10
  Speaker 1: to happen when the Milky Way collides with Andromeda or
• 26:12
  Speaker 1: the far future of our universe. Simulations give you that power.
• 26:16
  Speaker 1: You don't have to sit around and wait for the
• 26:18
  Speaker 1: events to play out. We can test it in simulation.
• 26:21
  Speaker 2: Right. That's pretty cool. And so what are the different
• 26:24
  Speaker 2: types of simulations that scientists use.
• 26:26
  Speaker 1: I would say that the simulation is almost everywhere in science.
• 26:29
  Speaker 1: You know, it used to be limited to a few
• 26:31
  Speaker 1: computationally complex fields, but now everybody sees how useful it is.
• 26:36
  Speaker 1: You know, even big companies like you want to design
• 26:39
  Speaker 1: a new airplane and you're considering a few different wing shapes.
• 26:42
  Speaker 1: It used to be you have to build prototypes of
• 26:44
  Speaker 1: those wing shapes and put them in a real huge
• 26:46
  Speaker 1: wind tunnel, very time consuming expensive. Now you can just
• 26:50
  Speaker 1: simulate the wind tunnel and get an idea for which
• 26:53
  Speaker 1: wing shape is going to work. You can explore thousands
• 26:55
  Speaker 1: of different shapes simultaneously, so that can be very very powerful.
• 27:00
  Speaker 1: So I think simulations are essentially everywhere in science now.
• 27:03
  Speaker 2: Well, I think they've been using simulations in things like
• 27:05
  Speaker 2: aerospace for a long time, right, Like even I'm thinking
• 27:08
  Speaker 2: in the space program in the fifties and sixties. I mean,
• 27:11
  Speaker 2: they didn't use physical computers, but they use people computers
• 27:14
  Speaker 2: to sort of simulate what the trajectories of the spacecraft
• 27:17
  Speaker 2: were going to be, right.
• 27:18
  Speaker 1: They definitely used human brains to do those calculations. Whether
• 27:22
  Speaker 1: you consider that a simulation, I think as a tricky point.
• 27:25
  Speaker 1: Is that just a theoretical calculation which people have been doing,
• 27:29
  Speaker 1: you know since Galileo or Francis Bacon or whatever. Is
• 27:32
  Speaker 1: it actually a simulation?
• 27:34
  Speaker 2: I don't know that. That's a tough question, all right,
• 27:35
  Speaker 2: So then what are some of the other types of
• 27:37
  Speaker 2: simulations people do, or what are some other ways that
• 27:40
  Speaker 2: physicists use simulations.
• 27:42
  Speaker 1: Another way they use them is to observe things that
• 27:45
  Speaker 1: they otherwise couldn't see, Like we want to know what's
• 27:48
  Speaker 1: going on inside the sun. Well, we have really no
• 27:50
  Speaker 1: prospects for actually seeing what's going on inside the Sun.
• 27:54
  Speaker 1: But we can build a simulation of the inside of
• 27:56
  Speaker 1: the Sun, and that's going to make predictions for things
• 27:59
  Speaker 1: that we can see, things happening on the surface of
• 28:01
  Speaker 1: the Sun, or the number of neutrinos coming to Earth,
• 28:04
  Speaker 1: and that helps us get an understanding for what's really
• 28:06
  Speaker 1: happening inside the Sun. And in the simulation, you're not limited, right,
• 28:10
  Speaker 1: you can ask questions about anything that's happening, like what
• 28:13
  Speaker 1: is the temperature or the core of the sun, what
• 28:15
  Speaker 1: is the velocity of the plasma. So often simulations always
• 28:18
  Speaker 1: checked by real experiments in places where we can observe
• 28:21
  Speaker 1: them give us access to things that we can't otherwise observe.
• 28:24
  Speaker 2: I think that's a crucial step in this process, right,
• 28:27
  Speaker 2: Like you can come up with this imaginary world in
• 28:29
  Speaker 2: your theater, but it has to match sort of what
• 28:31
  Speaker 2: you see at the end with reality, right absolutely.
• 28:34
  Speaker 1: Otherwise it's just science fiction, which you know has its
• 28:37
  Speaker 1: own value. But there is a special interest in our
• 28:40
  Speaker 1: universe and that's led to all sorts of deep understanding.
• 28:43
  Speaker 1: You know, the original simulations of the Sun predicted a
• 28:45
  Speaker 1: huge number of neutrinos landing on the surface of the Earth,
• 28:48
  Speaker 1: and they went out and measured them and the answer
• 28:50
  Speaker 1: was wrong, and they thought, did we get the sun wrong?
• 28:52
  Speaker 1: Or is there something going on with neutrinos? And it
• 28:55
  Speaker 1: turned out the simulation of the sun was correct and
• 28:57
  Speaker 1: neutrinos were doing something wonky between there and.
• 29:00
  Speaker 2: Here, right. I think the idea is that, you know,
• 29:01
  Speaker 2: if you create a simulation and you tweak the parameters
• 29:04
  Speaker 2: of it, right, like the numbers in it, so that
• 29:07
  Speaker 2: it matches what you see coming, for example, out of
• 29:09
  Speaker 2: the real Sun, then the idea is that maybe what
• 29:13
  Speaker 2: do you think is going on inside the sun is
• 29:15
  Speaker 2: actually what is going on inside the sun?
• 29:17
  Speaker 1: Yeah, that's exactly right. And somebody else might come up
• 29:20
  Speaker 1: with another simulation saying, actually, I think something else is happening.
• 29:24
  Speaker 1: And then you can ask, well, what's the difference between
• 29:26
  Speaker 1: these two simulations. Do they predict any different things that
• 29:29
  Speaker 1: we actually can measure that you can go off and
• 29:32
  Speaker 1: use that to distinguish between two various ideas. And we
• 29:35
  Speaker 1: talk about this all the time on the podcast. Sometimes
• 29:37
  Speaker 1: we have like two different possible ideas for what's happening
• 29:40
  Speaker 1: in near black holes. Remember we once talked about the
• 29:43
  Speaker 1: magnetic field near black holes something we could definitely not measure.
• 29:47
  Speaker 1: And there were two different models. One was called mad,
• 29:49
  Speaker 1: one was called sane, and they made slightly different predictions.
• 29:52
  Speaker 1: And then the recent picture of the black hole helped
• 29:55
  Speaker 1: us distinguish between these two models, these two simulations for
• 29:59
  Speaker 1: black hole magnet fields.
• 30:00
  Speaker 2: Right. But I guess that's the tricky thing, is like
• 30:03
  Speaker 2: just because the simulation matches what you see at the end,
• 30:05
  Speaker 2: it may not necessarily be what's going on inside, Right,
• 30:08
  Speaker 2: It could just be sort of a coincidence that it matches.
• 30:11
  Speaker 1: It certainly could be, And you always have to be
• 30:13
  Speaker 1: careful trusting your simulation. You always need ways to validate
• 30:16
  Speaker 1: it and to ensure that the bits that are important
• 30:18
  Speaker 1: to your science question are accurate.
• 30:20
  Speaker 2: Because I guess sometimes that's the only option that we have, right,
• 30:24
  Speaker 2: Like you're saying you can't just stick a stick inside
• 30:27
  Speaker 2: the science and see what's going on and things like
• 30:30
  Speaker 2: maybe black holes or the Big Bang, Like there's no
• 30:32
  Speaker 2: way where we can go back in time and do
• 30:34
  Speaker 2: an experiment on the Big Bang, Right, So we sort
• 30:36
  Speaker 2: of have to rely on these simulations to try to
• 30:39
  Speaker 2: understand what was going on.
• 30:41
  Speaker 1: Yeah, And they've turned out to be extraordinarily powerful tools
• 30:43
  Speaker 1: that give us insight into what might have happened in
• 30:45
  Speaker 1: the early universe or what's going on in the hearts
• 30:47
  Speaker 1: of black holes or neutron stars. I can't really imagine
• 30:50
  Speaker 1: doing science without them.
• 30:52
  Speaker 2: All right, Well, that's sort of what I guess a
• 30:54
  Speaker 2: pretty good answer for why scientists use simulations. And surprise, twist,
• 30:58
  Speaker 2: this whole conversation was just a simulation of our discussion
• 31:02
  Speaker 2: of the topic. This is like a sixth sense. I
• 31:04
  Speaker 2: only see simulated people. This was not the real podcast, right, Daniel.
• 31:08
  Speaker 1: That's right? Yeah, and hopefully this answer is also true
• 31:11
  Speaker 1: in the real universe.
• 31:12
  Speaker 2: But Daniel, you got to interview a scientist who does
• 31:15
  Speaker 2: physics and actually also wrote a book about simulating things
• 31:19
  Speaker 2: in the universe.
• 31:20
  Speaker 1: That's right. I had a fun chat with Professor Andrew Pnsen.
• 31:24
  Speaker 1: He's a cosmologist and a professor at University of College London,
• 31:27
  Speaker 1: and he wrote a new fun book called Universe in
• 31:30
  Speaker 1: a Box, which explores the role of simulation in cosmology
• 31:34
  Speaker 1: and in science in general. And he does have an
• 31:36
  Speaker 1: answer for the question is our universe a simulation?
• 31:40
  Speaker 2: Now? If I order Universe in a Box, do I
• 31:42
  Speaker 2: get a universe in a box.
• 31:45
  Speaker 1: Maybe you get a recipe for how to put a
• 31:47
  Speaker 1: universe into a box.
• 31:48
  Speaker 2: That's a that's not the universe in a box? And
• 31:51
  Speaker 2: also why a box? Why not? I don't know, as
• 31:55
  Speaker 2: spherical container.
• 31:56
  Speaker 1: I thought you were going to ask, if the universe
• 31:57
  Speaker 1: is in a box, what universe is the box in?
• 32:00
  Speaker 2: Hmmm, No, I wasn't going to ask that.
• 32:05
  Speaker 1: Dang it will my simulated Jorge wind Riven.
• 32:07
  Speaker 2: It's in the multiverse. I don't know. Aren't they like
• 32:10
  Speaker 2: meta universes outside of our universe? Isn't that the idea?
• 32:13
  Speaker 1: Click on the multiverse box option on Amazon Shipping.
• 32:16
  Speaker 2: The question then is can you have a multiverse in
• 32:19
  Speaker 2: a box? Anyways, you had a great conversation with Andrew.
• 32:22
  Speaker 1: I certainly did.
• 32:23
  Speaker 2: What motivated him to write this book.
• 32:25
  Speaker 1: You felt like the role of simulation in science was
• 32:27
  Speaker 1: super important and yet hadn't really been explored by any
• 32:31
  Speaker 1: pop side book, and so he wanted to share his
• 32:33
  Speaker 1: love for simulations with everybody.
• 32:36
  Speaker 2: Cool. Well, here is Daniel's interview with Professor Andrew Ponson
• 32:39
  Speaker 2: about his new book, Universe in a Box.
• 32:44
  Speaker 1: So then it's my pleasure to welcome to the program.
• 32:47
  Speaker 1: Andrew Ponson, a cosmologist and professor at University College London,
• 32:52
  Speaker 1: and youw thank you very much for joining us today.
• 32:54
  Speaker 3: Oh, thank you.
• 32:55
  Speaker 1: So your book is called Universe in a Box. It's
• 32:59
  Speaker 1: a fascinating and telling history and sort of definition of
• 33:02
  Speaker 1: what is a simulation and why it's important in science.
• 33:06
  Speaker 1: So let's start off with very very basics. What is
• 33:09
  Speaker 1: a simulation in your point of view?
• 33:11
  Speaker 6: There are different definitions you can give, but I think
• 33:13
  Speaker 6: a good place to start is by thinking it's trying
• 33:16
  Speaker 6: to capture some element of the real world inside a computer,
• 33:21
  Speaker 6: and that can take many different forms. It doesn't even
• 33:24
  Speaker 6: have to be a physics right, it's we can have
• 33:27
  Speaker 6: simulations of something like human behavior. There are simulations of
• 33:31
  Speaker 6: the way that crowds might behave that architects use to
• 33:35
  Speaker 6: make safer buildings by making the passageways the right kind
• 33:38
  Speaker 6: of size and shape that if there's an emergency situation,
• 33:41
  Speaker 6: then humans will evacuate the building in the most efficient,
• 33:45
  Speaker 6: safe way. But I think what that already teaches you
• 33:48
  Speaker 6: is that it's possible to do a simulation of something
• 33:51
  Speaker 6: without necessarily understanding everything about that thing before you start.
• 33:56
  Speaker 6: Because if you think of crowds, they're made out of people.
• 34:00
  Speaker 6: We can't actually predict everything about how an individual human's
• 34:04
  Speaker 6: going to react in any given scenario, And yet we
• 34:08
  Speaker 6: can make simulations that are useful. They might not be perfect,
• 34:12
  Speaker 6: but they're useful for giving us some insight into the
• 34:15
  Speaker 6: way that crowds might behave. So when you take that
• 34:18
  Speaker 6: across to the physics environment, and in particular my area cosmology,
• 34:23
  Speaker 6: it's about trying to capture something about how the universe behaves.
• 34:28
  Speaker 6: But we know from the outset we're never going to
• 34:30
  Speaker 6: get that perfect.
• 34:31
  Speaker 1: So then where is the value in a simulation if
• 34:34
  Speaker 1: you have to, like encode in already what's going to
• 34:37
  Speaker 1: happen when people bump into each other, or the purchasing
• 34:41
  Speaker 1: choices of people, or how galaxies interact. If you have
• 34:44
  Speaker 1: to already build in the physics, what are you learning
• 34:47
  Speaker 1: from the simulation? How do you get any information out
• 34:49
  Speaker 1: of it?
• 34:50
  Speaker 6: Well, the point is that you code in some things
• 34:53
  Speaker 6: about how the individual bits within your simulation behave. I
• 34:57
  Speaker 6: guess you know, in the case of the crowd that
• 34:59
  Speaker 6: would be how an individual human might behave under a
• 35:03
  Speaker 6: variety of circumstances. But in the case of physics, it
• 35:06
  Speaker 6: might be how we think dark matter particles flow through
• 35:10
  Speaker 6: the universe and interact with each other through gravity, and
• 35:13
  Speaker 6: what the simulation does is take a very large number
• 35:17
  Speaker 6: of those elements and kind of have them all individually
• 35:21
  Speaker 6: doing their thing. But the behavior that then emerges can
• 35:25
  Speaker 6: be very hard to anticipate in advance. And that's the point.
• 35:29
  Speaker 6: Understanding how a crowd behaves is not at all the
• 35:33
  Speaker 6: same thing as understanding how a human behaves, and in
• 35:36
  Speaker 6: the same way, understanding how dark matter behaves through our
• 35:40
  Speaker 6: whole cosmos is not at all the same thing as
• 35:43
  Speaker 6: understanding what an individual particle of dark matter might do.
• 35:47
  Speaker 1: This principle of emergent phenomena is something I'm super fascinated by.
• 35:51
  Speaker 1: It's incredible to me that sometimes we have laws of
• 35:54
  Speaker 1: physics at one scale, which you know, causally determine different
• 35:58
  Speaker 1: sort of laws of physics at another scale, which we
• 36:01
  Speaker 1: can't always easily predict. But as you say, we can
• 36:04
  Speaker 1: observe in action if we can, you know, construct the
• 36:07
  Speaker 1: right setup to you is simulation. Is it a kind
• 36:10
  Speaker 1: of experiment, Is it a kind of theory, or is
• 36:13
  Speaker 1: it sort of a new branch of science.
• 36:16
  Speaker 6: I think it's a new branch, but I think it's
• 36:17
  Speaker 6: got something in common with theory and with experiments, and
• 36:20
  Speaker 6: I guess I tilt mainly towards thinking of it as
• 36:23
  Speaker 6: an experiment. Now that's a little bit controversial. Sometimes people say, well,
• 36:28
  Speaker 6: it can't be an experiment. You've told the computer what
• 36:31
  Speaker 6: to do, Whereas in an experiment, you're supposed to go
• 36:34
  Speaker 6: and ask nature. You know, you're supposed to put things
• 36:37
  Speaker 6: to the test and confront them with the reality of
• 36:40
  Speaker 6: how things really work.
• 36:41
  Speaker 3: But you know, I'm not.
• 36:43
  Speaker 6: Sure that that distinction is always so clear. So an
• 36:46
  Speaker 6: example that I give in the book is, let's say
• 36:49
  Speaker 6: you're just trying to build an aircraft and you have
• 36:53
  Speaker 6: some idea about how you want to shape the wing,
• 36:56
  Speaker 6: but you don't know exactly how that wing is going
• 36:58
  Speaker 6: to perform. Now you now have a choice. You could
• 37:01
  Speaker 6: build a scale model of your wing and put it
• 37:05
  Speaker 6: inside a wind tunnel and see how it performs inside
• 37:08
  Speaker 6: a wind tunnel, and that's kind of an experiment. Or
• 37:11
  Speaker 6: you could make a digital version of your wing and
• 37:15
  Speaker 6: put it inside an airflow inside a computer that's a
• 37:19
  Speaker 6: kind of simulated airflow, and see how it behaves there.
• 37:22
  Speaker 6: And both of those are going to have limitations. There's
• 37:25
  Speaker 6: definitely limitations on what you can achieve inside the computer,
• 37:28
  Speaker 6: but there's also limitations in what you can achieve in
• 37:31
  Speaker 6: a wind tunnel. You can't make an infinitely big wind tunnel.
• 37:35
  Speaker 6: It's going to have edges, things are going to be
• 37:37
  Speaker 6: the wrong scale. So when you do experiments, you are
• 37:42
  Speaker 6: making some set of assumptions about the real world and
• 37:46
  Speaker 6: how what you're doing applies to the real world. And
• 37:49
  Speaker 6: I think that's just that's true in simulations as well.
• 37:51
  Speaker 6: So overall, this is why I start to think more
• 37:54
  Speaker 6: and more of simulations as types of experiment.
• 37:57
  Speaker 1: I have an argument with my brother who's a computer scientist,
• 38:00
  Speaker 1: and he runs what he calls experiments on his machine
• 38:04
  Speaker 1: learning models. I'm like, that's not an experiment. You're just
• 38:07
  Speaker 1: doing it in your computer. But you're absolutely right that
• 38:10
  Speaker 1: if you don't know the outcome and you're learning something,
• 38:12
  Speaker 1: it can be considered also an experiment. But you mentioned
• 38:15
  Speaker 1: something which I wanted to ask you about anyway, which
• 38:17
  Speaker 1: is the limitation of simulation. You're concocting sort of an
• 38:21
  Speaker 1: artificial universe, and you're learning something about that universe. If
• 38:25
  Speaker 1: that universe doesn't follow the same rules as ours, then
• 38:28
  Speaker 1: obviously we're not learning something about reality, which usually is
• 38:31
  Speaker 1: the goal. Can you say something about how we know
• 38:35
  Speaker 1: when to trust our simulations with the fundamental limitations of simulation.
• 38:39
  Speaker 6: Are yeah, I mean that is the hardest, most difficult question.
• 38:42
  Speaker 6: At the heart of doing good simulation is knowing what
• 38:45
  Speaker 6: to trust and what not to trust, and often it's hard,
• 38:49
  Speaker 6: you know, it can be really hard to know. I mean, fundamentally,
• 38:52
  Speaker 6: the limitation is just computational power that even if you're
• 38:57
  Speaker 6: doing something like a weather forecast, which I talk about
• 38:59
  Speaker 6: a bit in the book, you know, just the atmosphere
• 39:03
  Speaker 6: of the Earth has so many molecules in it that
• 39:06
  Speaker 6: you're never going to track each individual molecule, right, So
• 39:09
  Speaker 6: you're going to have to make some kind of approximation.
• 39:12
  Speaker 6: You're going to parcel up the air into almost like
• 39:15
  Speaker 6: big hypothetical bags of air that move around through our
• 39:18
  Speaker 6: atmosphere and use some laws to describe that. But then
• 39:23
  Speaker 6: you're going to have to go and say, well, you know,
• 39:25
  Speaker 6: we're not getting all the small scale details right, We're
• 39:28
  Speaker 6: going to have to put in some corrections. In the
• 39:31
  Speaker 6: case of meteorology, even clouds can be quite hard to
• 39:35
  Speaker 6: predict because you're just not getting all of those tiny
• 39:38
  Speaker 6: details that contribute to the way that a cloud forms
• 39:42
  Speaker 6: in reality. So you need to go in and put
• 39:44
  Speaker 6: into your simulation some kind of correction almost by hand.
• 39:48
  Speaker 6: You say, under these circumstances, clouds must start to form.
• 39:53
  Speaker 6: And you know, if you're a weather forecaster, you see
• 39:56
  Speaker 6: how well did I do by making that assumption about
• 40:00
  Speaker 6: how clouds form? And over time you sort of incrementally
• 40:03
  Speaker 6: improve by comparing how your simulation did with the reality
• 40:07
  Speaker 6: of how the weather unfolded. So we can do something
• 40:11
  Speaker 6: a bit similar in cosmology. It's not quite the same
• 40:13
  Speaker 6: because we don't get to kind of do the repeat
• 40:16
  Speaker 6: experiments in quite the same way as you do in
• 40:18
  Speaker 6: weather forecasting, for example. In some level, it's the same
• 40:21
  Speaker 6: kind of iterative process that we're getting better over time.
• 40:25
  Speaker 6: We're understanding the way that we have to put in
• 40:28
  Speaker 6: corrections to our simulations to account for things like the
• 40:32
  Speaker 6: way that stars evolve and change over time and dump
• 40:36
  Speaker 6: energy into the universe, and what black holes are up to,
• 40:39
  Speaker 6: and all of these things that we actually have to
• 40:41
  Speaker 6: help the computer along the way, if you like.
• 40:44
  Speaker 1: So why is it that we need to make these corrections?
• 40:47
  Speaker 1: We have these limitations is it purely just computational power.
• 40:51
  Speaker 1: In the limit of infinite computing power, could we predict
• 40:55
  Speaker 1: the weather tomorrow starting from particle physics and modeling every
• 40:58
  Speaker 1: single quirk in the atmosphe sphere, or is there a
• 41:01
  Speaker 1: conceptual limit there some obstacle that we can't overcome even
• 41:04
  Speaker 1: with infinite computing.
• 41:06
  Speaker 3: I think both are a problem.
• 41:07
  Speaker 6: So, first of all, we are very far from having
• 41:10
  Speaker 6: infinite computer power, a very very long way away from that.
• 41:13
  Speaker 6: But secondly, you're right, I mean there are more fundamental
• 41:16
  Speaker 6: limitations as well. In particular, we do not know exactly
• 41:20
  Speaker 6: where every molecule is in the atmosphere to start with.
• 41:24
  Speaker 6: So even if you had a computer powerful enough to
• 41:27
  Speaker 6: track at the molecular level what our atmosphere is doing,
• 41:31
  Speaker 6: you wouldn't know how to start the simulation. You wouldn't
• 41:34
  Speaker 6: have enough data to tell it what the atmosphere looks
• 41:38
  Speaker 6: like today. So there's an inaccuracy that's sort of just
• 41:41
  Speaker 6: coming from not having that perfect data, and an effect
• 41:45
  Speaker 6: known as chaos means that imperfect initial data very quickly
• 41:50
  Speaker 6: turns into big errors. So there's a kind of famous
• 41:55
  Speaker 6: example of this. It was Edward Lorenz who sort of
• 41:58
  Speaker 6: gave the thought experiment of butterfly flapping its wings somewhere
• 42:03
  Speaker 6: in Europe, say, and it has a sort of series
• 42:06
  Speaker 6: of knock on effects that over time just amplify and amplify,
• 42:09
  Speaker 6: and eventually the tiny little gust from the butterfly's wings
• 42:14
  Speaker 6: actually stimulates the formation of a hurricane. And you know,
• 42:19
  Speaker 6: these kind of effects, we know they're there, and physics
• 42:21
  Speaker 6: we call them chaos. And so you know, the slightest
• 42:26
  Speaker 6: inaccuracy in how you set things up will eventually make
• 42:31
  Speaker 6: the simulation depart from reality.
• 42:34
  Speaker 3: So we know that's true in cosmology as well.
• 42:36
  Speaker 6: And we don't have, you know, perfect information about the
• 42:40
  Speaker 6: early universe. We have quite good ideas for what was
• 42:43
  Speaker 6: going on there, but it's imperfect, and that means because
• 42:48
  Speaker 6: of chaos and the way that those imperfections are amplified
• 42:51
  Speaker 6: over time, what we end up with is, in some sense,
• 42:55
  Speaker 6: it's like a statistical.
• 42:56
  Speaker 3: Recreation of the universe.
• 42:59
  Speaker 6: It's telling us to hisstically what sorts of things should
• 43:03
  Speaker 6: be in the universe and what sorts of mixtures and
• 43:05
  Speaker 6: what kind of patterns, rather than literally recreating the universe.
• 43:10
  Speaker 6: So it's almost more like sort of climates, like a
• 43:14
  Speaker 6: climate simulation almost rather than a weather simulation.
• 43:18
  Speaker 1: I'm very interested in the history of simulations as well.
• 43:20
  Speaker 1: I mean, theoretical science is like thousands of years old.
• 43:24
  Speaker 1: Experimental science people argue about might be hundreds of years old.
• 43:27
  Speaker 1: Simulation based science seems like decades old. Can you take
• 43:30
  Speaker 1: us back to the root of it? Where does it begin? Really?
• 43:33
  Speaker 4: Well?
• 43:33
  Speaker 6: I think you know, the very earliest route you can
• 43:35
  Speaker 6: find is in the nineteenth century, where Charles Babbage and
• 43:39
  Speaker 6: Ada Lovelace were working on the idea of a computer
• 43:43
  Speaker 6: very similar to our modern computers. It was the first
• 43:46
  Speaker 6: time really anybody expressed the idea of having a machine
• 43:49
  Speaker 6: that could be told to perform any calculation. So before
• 43:54
  Speaker 6: that there were machines that did specific calculations, but this
• 43:58
  Speaker 6: was the first time somebody envisage a machine where you
• 44:01
  Speaker 6: could just give it instructions and it would carry out
• 44:04
  Speaker 6: calculations to your specifications. And Ada Lovelace actually wrote at
• 44:09
  Speaker 6: that time that one of the applications of being able
• 44:12
  Speaker 6: to do that was to be able to take what
• 44:15
  Speaker 6: we think are the governing laws of our physics and
• 44:19
  Speaker 6: make them kind of practical, you know, get the computer
• 44:23
  Speaker 6: to do all of the calculations that turns those abstract
• 44:27
  Speaker 6: equations into concrete, specific predictions for different scenarios.
• 44:33
  Speaker 3: So that's probably the first.
• 44:34
  Speaker 6: Time anyone expressed what we would recognize as a modern simulation. Then,
• 44:40
  Speaker 6: in terms of actually performing simulations, remarkably, some people tried
• 44:44
  Speaker 6: to do this in the twentieth century before digital computers
• 44:49
  Speaker 6: were actually made. So there are some beautiful stories like
• 44:56
  Speaker 6: a crazy character called Lewis Fry Richardson who was actually
• 45:00
  Speaker 6: on the front line of World War One trying to
• 45:03
  Speaker 6: calculate weather forecasts using pen and paper, very very repetitive
• 45:09
  Speaker 6: calculations he was doing that would be exactly what a
• 45:12
  Speaker 6: computer does today to do a weather forecast, but he
• 45:16
  Speaker 6: was doing it just with pen and paper and taking him,
• 45:19
  Speaker 6: you know, weeks stretching out into years just to do
• 45:22
  Speaker 6: one forecast. He wasn't trying to be practical about it.
• 45:25
  Speaker 6: He was just trying to prove a point that this
• 45:27
  Speaker 6: is actually doable in practice. And then you know, by
• 45:31
  Speaker 6: the end of World War Two there were actual computers available,
• 45:35
  Speaker 6: and very quickly from there the whole business of simulating
• 45:39
  Speaker 6: all sorts of different things, but ultimately the entire universe
• 45:43
  Speaker 6: kind of grew up quite quickly from there.
• 45:57
  Speaker 1: Tell us more about the role of simulation in or
• 46:00
  Speaker 1: personal research. Is this something you explore because you're fascinated
• 46:03
  Speaker 1: by the computer science of it, or to use it
• 46:05
  Speaker 1: just a tool that helps answer your physics questions.
• 46:08
  Speaker 6: I think it depends on the day you ask me.
• 46:10
  Speaker 6: I mean, some days I really enjoy the computer science
• 46:14
  Speaker 6: of all this, and you know, it's undeniably cool to
• 46:17
  Speaker 6: get to work with some of the world's biggest supercomputers
• 46:21
  Speaker 6: and be able to instruct them to carry out these
• 46:24
  Speaker 6: kind of simulations, and the results are enormous fun to
• 46:28
  Speaker 6: work with as well. So there is a bit of
• 46:30
  Speaker 6: there's a bit of that kind of nerdery in it.
• 46:33
  Speaker 6: But I think ultimately the thing that really keeps me
• 46:36
  Speaker 6: hooked is the idea that we are contributing to a
• 46:40
  Speaker 6: bigger picture of how our universe evolved, of the role
• 46:44
  Speaker 6: for materials in it that we as yet don't understand.
• 46:49
  Speaker 3: Things like dark matter and dark.
• 46:51
  Speaker 6: Energy that we know they're out there, they seem to
• 46:54
  Speaker 6: be having a profound effect on our universe, but we
• 46:57
  Speaker 6: really don't know what they are, and we're trying to
• 46:59
  Speaker 6: learn more about that. And then I suppose ultimately what
• 47:03
  Speaker 6: we're building towards is a better understanding of where we
• 47:06
  Speaker 6: came from. You know, the existence of us carbon based
• 47:10
  Speaker 6: life forms on this rocky planet is part of the
• 47:14
  Speaker 6: story that we're telling. Because the chemical elements from which
• 47:17
  Speaker 6: our planet and life are constructed weren't there in the
• 47:21
  Speaker 6: Big Bang. They've been manufactured over time. They need very
• 47:25
  Speaker 6: specific conditions to be manufactured and then concentrated enough to
• 47:30
  Speaker 6: start forming planets and enabling life and so on. So
• 47:34
  Speaker 6: I think, you know, ultimately that's the thing I'm most
• 47:37
  Speaker 6: excited by that we are telling this bigger story that
• 47:41
  Speaker 6: in the end speaks to a kind of deep question
• 47:43
  Speaker 6: within all of us about where did we come from?
• 47:46
  Speaker 1: Have you seen yet machine learning being used to amplify
• 47:50
  Speaker 1: or speed up, or just overall enhance simulations in your research.
• 47:55
  Speaker 6: Yeah, I mean machine learning is more and more important
• 47:58
  Speaker 6: throughout astrophysics. So it's being used in a variety of
• 48:02
  Speaker 6: different ways. One is to try and improve on some
• 48:06
  Speaker 6: of these things that we were talking about a moment
• 48:08
  Speaker 6: ago about you know, what do you do about the
• 48:12
  Speaker 6: things that you can't quite get right, Like the way
• 48:16
  Speaker 6: that stars form out of gas just such a complicated
• 48:20
  Speaker 6: process that we can't capture it perfectly the way that
• 48:23
  Speaker 6: those stars then put energy back into the galaxy that
• 48:27
  Speaker 6: they're forming within the roles of black holes, all of
• 48:30
  Speaker 6: these things that are very complicated and multifaceted.
• 48:35
  Speaker 3: We can use machine learning.
• 48:37
  Speaker 6: To do some of the hard work for us to
• 48:40
  Speaker 6: learn from examples of individual simulated stars, say about how
• 48:46
  Speaker 6: they behave, and kind of learn the lessons from those
• 48:49
  Speaker 6: and then take them and put them in the bigger
• 48:51
  Speaker 6: setting of trying to simulate then hundreds of billions of
• 48:55
  Speaker 6: stars across a galaxy or maybe you know, even out
• 48:58
  Speaker 6: into the universe. Sochine learning can kind of help us
• 49:02
  Speaker 6: take lessons from one bit of our simulations or one
• 49:06
  Speaker 6: bit of physics and insert them in an efficient way
• 49:10
  Speaker 6: into other simulations. That's one role for them, But they're
• 49:15
  Speaker 6: also crucial in interpreting the data we get from the
• 49:18
  Speaker 6: real universe. So the data that is coming from telescopes
• 49:24
  Speaker 6: and especially big new survey telescopes, things like EUCLID and
• 49:29
  Speaker 6: the Verra Rubin Observatory, these giant efforts to scan the
• 49:33
  Speaker 6: sky and build maps of our universe. They need machine
• 49:37
  Speaker 6: learning because the machine learning can kind of do a
• 49:41
  Speaker 6: lot of the initial data processing figuring out what's interesting,
• 49:46
  Speaker 6: what we're actually looking at, where it is in three
• 49:50
  Speaker 6: D space, and what needs to be flagged for further
• 49:54
  Speaker 6: human follow up. All of these things, machine learning is
• 49:57
  Speaker 6: playing an increasingly important role.
• 50:00
  Speaker 1: And in your view, what does a future hold for
• 50:02
  Speaker 1: simulation based science? Are there fields of science that don't
• 50:05
  Speaker 1: yet use any simulation and are on the cusp of
• 50:08
  Speaker 1: being revolutionized by this powerful new tool.
• 50:10
  Speaker 6: I mean, I'm not aware of fields that have shied
• 50:13
  Speaker 6: away from simulation. I think it is such a powerful
• 50:16
  Speaker 6: tool that when you start looking for it, you do
• 50:19
  Speaker 6: find it in use absolutely everywhere. But I think what
• 50:22
  Speaker 6: we can say about simulation is that we're still a
• 50:25
  Speaker 6: very early stage of understanding how to use simulation and
• 50:30
  Speaker 6: what roles it can play within the overall scientific progress. So,
• 50:35
  Speaker 6: you know, it goes all the way back to what
• 50:37
  Speaker 6: really is a simulation? Is it an experiment or is
• 50:39
  Speaker 6: it a calculation? How should we think about it? And
• 50:43
  Speaker 6: tied up with that is this question of how can
• 50:46
  Speaker 6: we improve our simulations? How do we get past that
• 50:49
  Speaker 6: stage of well, we just have to kind of play
• 50:53
  Speaker 6: around with things and tweak them till they fit because
• 50:56
  Speaker 6: of the intrinsic limitations. So I think we're at a
• 51:00
  Speaker 6: very early stage in understanding all of these things. So
• 51:04
  Speaker 6: for certain the role that simulations play is going to
• 51:07
  Speaker 6: change and evolve and I hope improve over the coming years.
• 51:12
  Speaker 1: So if you use the words universe and simulation together
• 51:15
  Speaker 1: in a sentence, that of course evokes in people's minds
• 51:18
  Speaker 1: this conversation that seems to be omnipresent, which is, you
• 51:21
  Speaker 1: know whether or not our universe could be a simulation
• 51:24
  Speaker 1: or the same question sort of when we build artificial
• 51:27
  Speaker 1: universes that have bits and pieces in it, could those
• 51:31
  Speaker 1: universes feel real to those occupants. So in your book,
• 51:34
  Speaker 1: you take a sort of skeptical view of the question
• 51:37
  Speaker 1: of whether we could be living in a simulation. Since
• 51:39
  Speaker 1: you're an expert on simulating universes, what are your arguments
• 51:42
  Speaker 1: against the concept that we could be living in a simulation.
• 51:45
  Speaker 6: Yeah, I think the primary argument is just to look
• 51:47
  Speaker 6: at the complexity of the universe that we're in. So
• 51:51
  Speaker 6: you can actually calculate something called the number of cubits,
• 51:54
  Speaker 6: so basically the number of quantum bits that you would
• 51:57
  Speaker 6: need in a quantum computer if you wanted to do
• 52:00
  Speaker 6: according to all the physics we know so far, if
• 52:03
  Speaker 6: you wanted to do a perfect simulation of our universe,
• 52:07
  Speaker 6: and that is a vast number. I forget it off
• 52:10
  Speaker 6: the top of my head. I think it's something like
• 52:11
  Speaker 6: ten to the one hundred and twenty four cubits.
• 52:13
  Speaker 3: It's something like that. It's a very very large number.
• 52:17
  Speaker 6: And right now, you know, we struggle even to make
• 52:19
  Speaker 6: a single cubit in a quantum computer. It's a vast
• 52:23
  Speaker 6: extrapolation from where we are now to the idea that
• 52:27
  Speaker 6: we will routinely be able to run simulations that have
• 52:31
  Speaker 6: the same kind of richness as the reality that we
• 52:35
  Speaker 6: currently live in.
• 52:36
  Speaker 3: But even more than that. If you ask, well, you know, what.
• 52:40
  Speaker 6: Resources would you need to build a computer that really
• 52:44
  Speaker 6: had that level of capability, Well, it turns out you
• 52:49
  Speaker 6: would need to make use of the entire universe just
• 52:53
  Speaker 6: to simulate one universe. Because physics puts limitations on information processing.
• 52:59
  Speaker 6: You can't process as much information as you like. There
• 53:02
  Speaker 6: are limitations placed on it according to the physical system
• 53:06
  Speaker 6: that it's being processed by, and so these come back
• 53:08
  Speaker 6: to bite you. So and that's what gives rise to
• 53:13
  Speaker 6: this claim that you can only simulate the whole universe
• 53:16
  Speaker 6: perfectly if you have access to the entire physical resources
• 53:20
  Speaker 6: of that universe. Now, there are lots of further objections
• 53:24
  Speaker 6: you can raise. You can say, well, what about, for example,
• 53:27
  Speaker 6: if the higher up universe where we're being simulated is
• 53:32
  Speaker 6: just a much bigger universe with many more cubits at
• 53:35
  Speaker 6: their disposal, and so these resources seem terribly trivial maybe
• 53:40
  Speaker 6: to the people in the higher up universe. But at
• 53:43
  Speaker 6: that point I kind of lose patience with the argument,
• 53:45
  Speaker 6: because it seems to me, at that point we it's
• 53:48
  Speaker 6: no longer a sort of obvious extrapolation from where we are.
• 53:52
  Speaker 6: Now you're now talking about hypothesizing beings with so much
• 53:56
  Speaker 6: more power and so much more technological capability than we have,
• 54:01
  Speaker 6: that we might as well just go and talk about religion,
• 54:04
  Speaker 6: because at this point it's lost contact, in my view,
• 54:07
  Speaker 6: with any science.
• 54:09
  Speaker 1: Wonderful. Well, I really enjoyed your book, And a question
• 54:12
  Speaker 1: I always have for folks who write popular science books
• 54:16
  Speaker 1: about very technical topics is why what do you think
• 54:20
  Speaker 1: that the general public, folks out there who are not
• 54:23
  Speaker 1: simulation experts need to know about simulation?
• 54:27
  Speaker 3: I think there were two reasons.
• 54:28
  Speaker 6: The first was it was amazing to me that nobody
• 54:31
  Speaker 6: had written about this topic before because it's so central
• 54:34
  Speaker 6: in our field of cosmology, and you know, cosmology is
• 54:38
  Speaker 6: something that people do talk about a lot. There's a
• 54:40
  Speaker 6: lot of it out there in the media and in books,
• 54:43
  Speaker 6: because it's genuinely you know, it's exciting, and I think
• 54:46
  Speaker 6: it speaks to all of us about where we came from.
• 54:49
  Speaker 6: And so it seems a real surprise to me that
• 54:51
  Speaker 6: this central tool in how we're learning about the universe
• 54:55
  Speaker 6: was not being written about, and so it felt to
• 54:57
  Speaker 6: me like it needs to be rectified. We need to
• 55:00
  Speaker 6: be talking about this because it offers immense strengths, you know,
• 55:04
  Speaker 6: real enormous strengths that I think are kind of hidden
• 55:08
  Speaker 6: away sometimes. But it also comes with lots of caveats,
• 55:12
  Speaker 6: some of them we've discussed. You know that we can't
• 55:15
  Speaker 6: do these perfect recreations of the universe. There are lots
• 55:18
  Speaker 6: of approximations. We might be making mistakes, and you know,
• 55:21
  Speaker 6: there's no way to rule that out. It's just part
• 55:23
  Speaker 6: of the scientific process that we have to keep an
• 55:26
  Speaker 6: open mind about these things. So I wanted to write
• 55:29
  Speaker 6: about that in a kind of open and honest way,
• 55:32
  Speaker 6: rather than sort of leaving the simulations as black boxes
• 55:36
  Speaker 6: that seemingly recreate our universe through some process of magic. No,
• 55:40
  Speaker 6: it's a very human process and it's got all of
• 55:44
  Speaker 6: the strengths and weaknesses that come with being that kind
• 55:47
  Speaker 6: of human process.
• 55:49
  Speaker 1: Wonderful. Well, thank you very much again. Professor Andrew Hanson
• 55:53
  Speaker 1: at University of College London and author of the book
• 55:55
  Speaker 1: The Universe in a Box Out now encourage everyone to
• 55:59
  Speaker 1: go out and about how science is actually done. Thanks
• 56:03
  Speaker 1: again very much Andrew for joining us today.
• 56:05
  Speaker 2: All Right, did you feel you had a real conversation
• 56:07
  Speaker 2: with him, Like, did things get real or was it
• 56:10
  Speaker 2: all just simulated pleasantries?
• 56:12
  Speaker 1: I was able to simulate enjoying a conversation. Yeah. Oh
• 56:17
  Speaker 1: that's sort of always my situation though, because I'm kind
• 56:20
  Speaker 1: of an introvert, so I have to simulate enjoying human interactions.
• 56:23
  Speaker 2: Yet you have to simulate being engaging human being.
• 56:27
  Speaker 1: I'm trying to quietly slip unnoticed into human society.
• 56:29
  Speaker 2: So you are a simulation, Daniel.
• 56:32
  Speaker 1: I'm simulating being a human.
• 56:34
  Speaker 2: But anyways, that was a great conversation. What's your main
• 56:37
  Speaker 2: takeaway from it?
• 56:37
  Speaker 1: To me, it's fascinating that so recently in the history
• 56:40
  Speaker 1: of science, just the last few decades, we've developed this
• 56:43
  Speaker 1: crucial new tool that we now think is indispensable. It
• 56:46
  Speaker 1: makes me wonder in fifty years and one hundred years,
• 56:49
  Speaker 1: what new branch of science or what new tool we're
• 56:52
  Speaker 1: going to add to our toolbox which future scientists will
• 56:55
  Speaker 1: think is indispensable, And we'll wonder, like, how did Daniel
• 56:58
  Speaker 1: and folks who lived way back then do any science
• 57:01
  Speaker 1: without it?
• 57:02
  Speaker 2: Or maybe even like why did Daniel do science? Why
• 57:05
  Speaker 2: didn't they choose use the AI physicists to answer all
• 57:09
  Speaker 2: the questions?
• 57:10
  Speaker 1: Yeah, stay in bed and let the AIS do the work.
• 57:12
  Speaker 1: The answer to that is the AIS wan't answer questions
• 57:14
  Speaker 1: the AIS are interested in and science is for people,
• 57:17
  Speaker 1: by people, and of people, So you got to have
• 57:20
  Speaker 1: people asking questions.
• 57:21
  Speaker 2: Wait, wait, do you mean the AIS won't always do
• 57:23
  Speaker 2: what we ask them to do.
• 57:25
  Speaker 1: They'll always do exactly what we ask them to do,
• 57:28
  Speaker 1: just not and maybe in the way that we want
• 57:30
  Speaker 1: them to.
• 57:31
  Speaker 2: All right, Well, stay tuned to see if we are
• 57:34
  Speaker 2: in a simulation right now? Maybe I guess some people
• 57:37
  Speaker 2: consider the whole universe to be service.
• 57:39
  Speaker 1: Simulation, right, how can we possibly know?
• 57:41
  Speaker 2: Yeah, Like, what's the difference if.
• 57:43
  Speaker 1: We are in a vast alien simulation. Then I definitely
• 57:46
  Speaker 1: want to talk to those aliens because they have some
• 57:48
  Speaker 1: awesome computers.
• 57:49
  Speaker 2: And also to stay tuned about what new developments humans
• 57:52
  Speaker 2: are going to discover using simulations and or artificial intelligence.
• 57:56
  Speaker 2: So we hope you enjoyed that. Thanks for joining us,
• 57:59
  Speaker 2: see you next time.
• 58:07
  Speaker 1: Thanks for listening, and remember that Daniel and Jorge Explain
• 58:10
  Speaker 1: the Universe is a production of iHeartRadio. For more podcasts
• 58:14
  Speaker 1: from iHeartRadio, visit the iHeartRadio app, Apple Podcasts, or wherever
• 58:19
  Speaker 1: you listen to your favorite shows.